Homomorphic Encryption for Fuzzy Data Matching

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current data processing technologies face challenges in efficiently handling encrypted data for similarity queries across networks, particularly due to high computational overheads and latency issues, which hinder effective data management and privacy protection in large-scale data environments.

Innovation Solution

A computer-implemented method using homomorphic encryption and probabilistic data structures, such as Bloom filters, to perform group-wise data similarity queries, reducing computational complexity and latency by aggregating data into matrices for efficient processing and matching operations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If homomorphic encryption is used to process encrypted data for similarity queries, then data privacy is protected, but computational overhead increases and turnaround time becomes prohibitively slow

Engineering Contradiction:
Improvedata privacy protectionVSAvoidcomputational efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent segments the encrypted data processing into distinct phases: encoding data into probabilistic data structures (Bloom filters), encrypting these structures using homomorphic encryption, and then performing similarity computations on the encrypted representations. This segmentation allows each phase to be optimized independently, balancing privacy protection with computational efficiency.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the parameter representation by transforming raw data into encoded forms (Bloom filter encodings) before encryption. This parameter transformation reduces the computational complexity of subsequent similarity operations while maintaining the ability to perform accurate matching on the encrypted data.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If traditional pair-wise matching is used for data similarity queries, then matching accuracy is maintained, but processing time and computational resources increase significantly

Engineering Contradiction:
Improvematching accuracyVSAvoidquery processing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent merges multiple individual data item comparisons into batch operations by organizing data into structured representations (e.g., matrices of Bloom filter encodings). This allows simultaneous processing of multiple data pairs, dramatically reducing the total number of operations required while maintaining matching accuracy through the properties of the underlying probabilistic data structures.

Inventive Principle:
Principle #5Merging (Combining)

3Reliability

If data is encrypted to protect privacy, then data security is improved, but data processing and matching operations cannot be performed

Engineering Contradiction:
Improvedata securityVSAvoiddata processing capability
Core Design Contradiction:
ReliabilityVSEase of operation

Solution Approach 1:

The patent introduces probabilistic data structures (Bloom filters) as an intermediary representation between the original data and the encrypted form. These structures serve as a mediator that preserves the essential matching properties of the data while enabling efficient homomorphic operations, thus bridging the gap between security requirements and processing capabilities.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP3935775B1Fuzzy datamatching using homomorphic encryption
Publication Date: 2024.02.14 BRAINLAB AG
  • EP3935775B1 patent drawingFigure 1
  • EP3935775B1 patent drawingFigure 2
  • EP3935775B1 patent drawingFigure 3

AI summary

Computer-implemented methods and systems (DMS) for facilitating data similarity queries across a network (CN) of data memories (DM1, DM2). The disclosed methods and system are configured for matching data items held in the data memories (DM1, DM2) in a probabilistic manner with cryptographic protection of the data items. The data matching methods and systems (DMS) are robust against inconsistencies within the data to be matched (such as typographical errors, minor mismatches etc), within a certain predetermined similarity threshold (q), usually described as a percentage. The disclosed methods and systems (DMS) allow fast, low latency turnaround by aggregating the data items to be matched into data structures (M1, M2) that facilitate group-wise matching as opposed to pair-wise matching.